Hi I have some further info about this. I have updated my version of brat and nothing changes for the real data. I don't think the window size has any effect on this and I use the same window myself with real data and have no problems in TPM2, T1, T2. It is harder to say for TPM1 becuase it is so messy. The problem in Djamels case is that only 4 clusters are found from the clustering, however there are also only 20 something TPC sequences when we looked. My suggestion is to print out the row numbers for all the TPC sequences and see if they are all there. If not : dig problem. Set Debug Level and look what goes wrong. If all rows are there and correct and min 2 sequence pr row : cluster problem - hard to beleive If the rows are not correct : geometry problem If there is only 1 sequence pr row : dig parameter problem - gain/n e- changed > I put a debug level in trackDummy (BrTrackFollowFinder) and noticed that > some local tracks do exist but it seems that they are not added to the > track tables (apparently) so that they remain empty in the event node. Not true. The track direction etc. you get with a high debug level are track segments that are eventually the tracks if #hits in the segment is high enough (Number of active rows - allowed to miss (3 is default I think). You should look at No. of track groups and number of local tracks. Track groups are groups of local tracks sharing many hits. Track groups are eventually reduced to one BrDetectorTrack pr group. Ciao Peter
This archive was generated by hypermail 2b29 : Fri Sep 08 2000 - 08:52:25 EDT